Merge branch 'linus' into timers/core
[safe/jmp/linux-2.6] / kernel / time / timekeeping.c
index e97c50f..caf8d4d 100644 (file)
 #include <linux/percpu.h>
 #include <linux/init.h>
 #include <linux/mm.h>
+#include <linux/sched.h>
 #include <linux/sysdev.h>
 #include <linux/clocksource.h>
 #include <linux/jiffies.h>
 #include <linux/time.h>
 #include <linux/tick.h>
+#include <linux/stop_machine.h>
+
+/* Structure holding internal timekeeping values. */
+struct timekeeper {
+       /* Current clocksource used for timekeeping. */
+       struct clocksource *clock;
+       /* The shift value of the current clocksource. */
+       int     shift;
+
+       /* Number of clock cycles in one NTP interval. */
+       cycle_t cycle_interval;
+       /* Number of clock shifted nano seconds in one NTP interval. */
+       u64     xtime_interval;
+       /* Raw nano seconds accumulated per NTP interval. */
+       u32     raw_interval;
+
+       /* Clock shifted nano seconds remainder not stored in xtime.tv_nsec. */
+       u64     xtime_nsec;
+       /* Difference between accumulated time and NTP time in ntp
+        * shifted nano seconds. */
+       s64     ntp_error;
+       /* Shift conversion between clock shifted nano seconds and
+        * ntp shifted nano seconds. */
+       int     ntp_error_shift;
+       /* NTP adjusted clock multiplier */
+       u32     mult;
+};
+
+struct timekeeper timekeeper;
+
+/**
+ * timekeeper_setup_internals - Set up internals to use clocksource clock.
+ *
+ * @clock:             Pointer to clocksource.
+ *
+ * Calculates a fixed cycle/nsec interval for a given clocksource/adjustment
+ * pair and interval request.
+ *
+ * Unless you're the timekeeping code, you should not be using this!
+ */
+static void timekeeper_setup_internals(struct clocksource *clock)
+{
+       cycle_t interval;
+       u64 tmp;
+
+       timekeeper.clock = clock;
+       clock->cycle_last = clock->read(clock);
+
+       /* Do the ns -> cycle conversion first, using original mult */
+       tmp = NTP_INTERVAL_LENGTH;
+       tmp <<= clock->shift;
+       tmp += clock->mult/2;
+       do_div(tmp, clock->mult);
+       if (tmp == 0)
+               tmp = 1;
+
+       interval = (cycle_t) tmp;
+       timekeeper.cycle_interval = interval;
 
+       /* Go back from cycles -> shifted ns */
+       timekeeper.xtime_interval = (u64) interval * clock->mult;
+       timekeeper.raw_interval =
+               ((u64) interval * clock->mult) >> clock->shift;
+
+       timekeeper.xtime_nsec = 0;
+       timekeeper.shift = clock->shift;
+
+       timekeeper.ntp_error = 0;
+       timekeeper.ntp_error_shift = NTP_SCALE_SHIFT - clock->shift;
+
+       /*
+        * The timekeeper keeps its own mult values for the currently
+        * active clocksource. These value will be adjusted via NTP
+        * to counteract clock drifting.
+        */
+       timekeeper.mult = clock->mult;
+}
+
+/* Timekeeper helper functions. */
+static inline s64 timekeeping_get_ns(void)
+{
+       cycle_t cycle_now, cycle_delta;
+       struct clocksource *clock;
+
+       /* read clocksource: */
+       clock = timekeeper.clock;
+       cycle_now = clock->read(clock);
+
+       /* calculate the delta since the last update_wall_time: */
+       cycle_delta = (cycle_now - clock->cycle_last) & clock->mask;
+
+       /* return delta convert to nanoseconds using ntp adjusted mult. */
+       return clocksource_cyc2ns(cycle_delta, timekeeper.mult,
+                                 timekeeper.shift);
+}
+
+static inline s64 timekeeping_get_ns_raw(void)
+{
+       cycle_t cycle_now, cycle_delta;
+       struct clocksource *clock;
+
+       /* read clocksource: */
+       clock = timekeeper.clock;
+       cycle_now = clock->read(clock);
+
+       /* calculate the delta since the last update_wall_time: */
+       cycle_delta = (cycle_now - clock->cycle_last) & clock->mask;
+
+       /* return delta convert to nanoseconds using ntp adjusted mult. */
+       return clocksource_cyc2ns(cycle_delta, clock->mult, clock->shift);
+}
 
 /*
  * This read-write spinlock protects us from races in SMP while
- * playing with xtime and avenrun.
+ * playing with xtime.
  */
 __cacheline_aligned_in_smp DEFINE_SEQLOCK(xtime_lock);
 
@@ -44,47 +155,54 @@ __cacheline_aligned_in_smp DEFINE_SEQLOCK(xtime_lock);
  */
 struct timespec xtime __attribute__ ((aligned (16)));
 struct timespec wall_to_monotonic __attribute__ ((aligned (16)));
-static unsigned long total_sleep_time;         /* seconds */
+static struct timespec total_sleep_time;
+
+/*
+ * The raw monotonic time for the CLOCK_MONOTONIC_RAW posix clock.
+ */
+struct timespec raw_time;
 
 /* flag for if timekeeping is suspended */
 int __read_mostly timekeeping_suspended;
 
-static struct timespec xtime_cache __attribute__ ((aligned (16)));
-void update_xtime_cache(u64 nsec)
+/* must hold xtime_lock */
+void timekeeping_leap_insert(int leapsecond)
 {
-       xtime_cache = xtime;
-       timespec_add_ns(&xtime_cache, nsec);
+       xtime.tv_sec += leapsecond;
+       wall_to_monotonic.tv_sec -= leapsecond;
+       update_vsyscall(&xtime, timekeeper.clock, timekeeper.mult);
 }
 
-struct clocksource *clock;
-
-
 #ifdef CONFIG_GENERIC_TIME
+
 /**
- * clocksource_forward_now - update clock to the current time
+ * timekeeping_forward_now - update clock to the current time
  *
  * Forward the current clock to update its state since the last call to
  * update_wall_time(). This is useful before significant clock changes,
  * as it avoids having to deal with this time offset explicitly.
  */
-static void clocksource_forward_now(void)
+static void timekeeping_forward_now(void)
 {
        cycle_t cycle_now, cycle_delta;
+       struct clocksource *clock;
        s64 nsec;
 
-       cycle_now = clocksource_read(clock);
+       clock = timekeeper.clock;
+       cycle_now = clock->read(clock);
        cycle_delta = (cycle_now - clock->cycle_last) & clock->mask;
        clock->cycle_last = cycle_now;
 
-       nsec = cyc2ns(clock, cycle_delta);
+       nsec = clocksource_cyc2ns(cycle_delta, timekeeper.mult,
+                                 timekeeper.shift);
 
        /* If arch requires, add in gettimeoffset() */
        nsec += arch_gettimeoffset();
 
        timespec_add_ns(&xtime, nsec);
 
-       nsec = ((s64)cycle_delta * clock->mult_orig) >> clock->shift;
-       clock->raw_time.tv_nsec += nsec;
+       nsec = clocksource_cyc2ns(cycle_delta, clock->mult, clock->shift);
+       timespec_add_ns(&raw_time, nsec);
 }
 
 /**
@@ -95,7 +213,6 @@ static void clocksource_forward_now(void)
  */
 void getnstimeofday(struct timespec *ts)
 {
-       cycle_t cycle_now, cycle_delta;
        unsigned long seq;
        s64 nsecs;
 
@@ -105,15 +222,7 @@ void getnstimeofday(struct timespec *ts)
                seq = read_seqbegin(&xtime_lock);
 
                *ts = xtime;
-
-               /* read clocksource: */
-               cycle_now = clocksource_read(clock);
-
-               /* calculate the delta since the last update_wall_time: */
-               cycle_delta = (cycle_now - clock->cycle_last) & clock->mask;
-
-               /* convert to nanoseconds: */
-               nsecs = cyc2ns(clock, cycle_delta);
+               nsecs = timekeeping_get_ns();
 
                /* If arch requires, add in gettimeoffset() */
                nsecs += arch_gettimeoffset();
@@ -125,6 +234,57 @@ void getnstimeofday(struct timespec *ts)
 
 EXPORT_SYMBOL(getnstimeofday);
 
+ktime_t ktime_get(void)
+{
+       unsigned int seq;
+       s64 secs, nsecs;
+
+       WARN_ON(timekeeping_suspended);
+
+       do {
+               seq = read_seqbegin(&xtime_lock);
+               secs = xtime.tv_sec + wall_to_monotonic.tv_sec;
+               nsecs = xtime.tv_nsec + wall_to_monotonic.tv_nsec;
+               nsecs += timekeeping_get_ns();
+
+       } while (read_seqretry(&xtime_lock, seq));
+       /*
+        * Use ktime_set/ktime_add_ns to create a proper ktime on
+        * 32-bit architectures without CONFIG_KTIME_SCALAR.
+        */
+       return ktime_add_ns(ktime_set(secs, 0), nsecs);
+}
+EXPORT_SYMBOL_GPL(ktime_get);
+
+/**
+ * ktime_get_ts - get the monotonic clock in timespec format
+ * @ts:                pointer to timespec variable
+ *
+ * The function calculates the monotonic clock from the realtime
+ * clock and the wall_to_monotonic offset and stores the result
+ * in normalized timespec format in the variable pointed to by @ts.
+ */
+void ktime_get_ts(struct timespec *ts)
+{
+       struct timespec tomono;
+       unsigned int seq;
+       s64 nsecs;
+
+       WARN_ON(timekeeping_suspended);
+
+       do {
+               seq = read_seqbegin(&xtime_lock);
+               *ts = xtime;
+               tomono = wall_to_monotonic;
+               nsecs = timekeeping_get_ns();
+
+       } while (read_seqretry(&xtime_lock, seq));
+
+       set_normalized_timespec(ts, ts->tv_sec + tomono.tv_sec,
+                               ts->tv_nsec + tomono.tv_nsec + nsecs);
+}
+EXPORT_SYMBOL_GPL(ktime_get_ts);
+
 /**
  * do_gettimeofday - Returns the time of day in a timeval
  * @tv:                pointer to the timeval to be set
@@ -157,7 +317,7 @@ int do_settimeofday(struct timespec *tv)
 
        write_seqlock_irqsave(&xtime_lock, flags);
 
-       clocksource_forward_now();
+       timekeeping_forward_now();
 
        ts_delta.tv_sec = tv->tv_sec - xtime.tv_sec;
        ts_delta.tv_nsec = tv->tv_nsec - xtime.tv_nsec;
@@ -165,12 +325,10 @@ int do_settimeofday(struct timespec *tv)
 
        xtime = *tv;
 
-       update_xtime_cache(0);
-
-       clock->error = 0;
+       timekeeper.ntp_error = 0;
        ntp_clear();
 
-       update_vsyscall(&xtime, clock);
+       update_vsyscall(&xtime, timekeeper.clock, timekeeper.mult);
 
        write_sequnlock_irqrestore(&xtime_lock, flags);
 
@@ -187,44 +345,97 @@ EXPORT_SYMBOL(do_settimeofday);
  *
  * Accumulates current time interval and initializes new clocksource
  */
-static void change_clocksource(void)
+static int change_clocksource(void *data)
 {
        struct clocksource *new, *old;
 
-       new = clocksource_get_next();
+       new = (struct clocksource *) data;
 
-       if (clock == new)
+       timekeeping_forward_now();
+       if (!new->enable || new->enable(new) == 0) {
+               old = timekeeper.clock;
+               timekeeper_setup_internals(new);
+               if (old->disable)
+                       old->disable(old);
+       }
+       return 0;
+}
+
+/**
+ * timekeeping_notify - Install a new clock source
+ * @clock:             pointer to the clock source
+ *
+ * This function is called from clocksource.c after a new, better clock
+ * source has been registered. The caller holds the clocksource_mutex.
+ */
+void timekeeping_notify(struct clocksource *clock)
+{
+       if (timekeeper.clock == clock)
                return;
+       stop_machine(change_clocksource, clock, NULL);
+       tick_clock_notify();
+}
 
-       clocksource_forward_now();
+#else /* GENERIC_TIME */
 
-       if (clocksource_enable(new))
-               return;
+static inline void timekeeping_forward_now(void) { }
 
-       new->raw_time = clock->raw_time;
-       old = clock;
-       clock = new;
-       clocksource_disable(old);
+/**
+ * ktime_get - get the monotonic time in ktime_t format
+ *
+ * returns the time in ktime_t format
+ */
+ktime_t ktime_get(void)
+{
+       struct timespec now;
 
-       clock->cycle_last = 0;
-       clock->cycle_last = clocksource_read(clock);
-       clock->error = 0;
-       clock->xtime_nsec = 0;
-       clocksource_calculate_interval(clock, NTP_INTERVAL_LENGTH);
+       ktime_get_ts(&now);
 
-       tick_clock_notify();
+       return timespec_to_ktime(now);
+}
+EXPORT_SYMBOL_GPL(ktime_get);
 
-       /*
-        * We're holding xtime lock and waking up klogd would deadlock
-        * us on enqueue.  So no printing!
-       printk(KERN_INFO "Time: %s clocksource has been installed.\n",
-              clock->name);
-        */
+/**
+ * ktime_get_ts - get the monotonic clock in timespec format
+ * @ts:                pointer to timespec variable
+ *
+ * The function calculates the monotonic clock from the realtime
+ * clock and the wall_to_monotonic offset and stores the result
+ * in normalized timespec format in the variable pointed to by @ts.
+ */
+void ktime_get_ts(struct timespec *ts)
+{
+       struct timespec tomono;
+       unsigned long seq;
+
+       do {
+               seq = read_seqbegin(&xtime_lock);
+               getnstimeofday(ts);
+               tomono = wall_to_monotonic;
+
+       } while (read_seqretry(&xtime_lock, seq));
+
+       set_normalized_timespec(ts, ts->tv_sec + tomono.tv_sec,
+                               ts->tv_nsec + tomono.tv_nsec);
 }
-#else
-static inline void clocksource_forward_now(void) { }
-static inline void change_clocksource(void) { }
-#endif
+EXPORT_SYMBOL_GPL(ktime_get_ts);
+
+#endif /* !GENERIC_TIME */
+
+/**
+ * ktime_get_real - get the real (wall-) time in ktime_t format
+ *
+ * returns the time in ktime_t format
+ */
+ktime_t ktime_get_real(void)
+{
+       struct timespec now;
+
+       getnstimeofday(&now);
+
+       return timespec_to_ktime(now);
+}
+EXPORT_SYMBOL_GPL(ktime_get_real);
 
 /**
  * getrawmonotonic - Returns the raw monotonic time in a timespec
@@ -236,21 +447,11 @@ void getrawmonotonic(struct timespec *ts)
 {
        unsigned long seq;
        s64 nsecs;
-       cycle_t cycle_now, cycle_delta;
 
        do {
                seq = read_seqbegin(&xtime_lock);
-
-               /* read clocksource: */
-               cycle_now = clocksource_read(clock);
-
-               /* calculate the delta since the last update_wall_time: */
-               cycle_delta = (cycle_now - clock->cycle_last) & clock->mask;
-
-               /* convert to nanoseconds: */
-               nsecs = ((s64)cycle_delta * clock->mult_orig) >> clock->shift;
-
-               *ts = clock->raw_time;
+               nsecs = timekeeping_get_ns_raw();
+               *ts = raw_time;
 
        } while (read_seqretry(&xtime_lock, seq));
 
@@ -270,7 +471,7 @@ int timekeeping_valid_for_hres(void)
        do {
                seq = read_seqbegin(&xtime_lock);
 
-               ret = clock->flags & CLOCK_SOURCE_VALID_FOR_HRES;
+               ret = timekeeper.clock->flags & CLOCK_SOURCE_VALID_FOR_HRES;
 
        } while (read_seqretry(&xtime_lock, seq));
 
@@ -278,17 +479,44 @@ int timekeeping_valid_for_hres(void)
 }
 
 /**
- * read_persistent_clock -  Return time in seconds from the persistent clock.
+ * timekeeping_max_deferment - Returns max time the clocksource can be deferred
+ *
+ * Caller must observe xtime_lock via read_seqbegin/read_seqretry to
+ * ensure that the clocksource does not change!
+ */
+u64 timekeeping_max_deferment(void)
+{
+       return timekeeper.clock->max_idle_ns;
+}
+
+/**
+ * read_persistent_clock -  Return time from the persistent clock.
  *
  * Weak dummy function for arches that do not yet support it.
- * Returns seconds from epoch using the battery backed persistent clock.
- * Returns zero if unsupported.
+ * Reads the time from the battery backed persistent clock.
+ * Returns a timespec with tv_sec=0 and tv_nsec=0 if unsupported.
  *
  *  XXX - Do be sure to remove it once all arches implement it.
  */
-unsigned long __attribute__((weak)) read_persistent_clock(void)
+void __attribute__((weak)) read_persistent_clock(struct timespec *ts)
 {
-       return 0;
+       ts->tv_sec = 0;
+       ts->tv_nsec = 0;
+}
+
+/**
+ * read_boot_clock -  Return time of the system start.
+ *
+ * Weak dummy function for arches that do not yet support it.
+ * Function to read the exact time the system has been started.
+ * Returns a timespec with tv_sec=0 and tv_nsec=0 if unsupported.
+ *
+ *  XXX - Do be sure to remove it once all arches implement it.
+ */
+void __attribute__((weak)) read_boot_clock(struct timespec *ts)
+{
+       ts->tv_sec = 0;
+       ts->tv_nsec = 0;
 }
 
 /*
@@ -296,29 +524,39 @@ unsigned long __attribute__((weak)) read_persistent_clock(void)
  */
 void __init timekeeping_init(void)
 {
+       struct clocksource *clock;
        unsigned long flags;
-       unsigned long sec = read_persistent_clock();
+       struct timespec now, boot;
+
+       read_persistent_clock(&now);
+       read_boot_clock(&boot);
 
        write_seqlock_irqsave(&xtime_lock, flags);
 
        ntp_init();
 
-       clock = clocksource_get_next();
-       clocksource_enable(clock);
-       clocksource_calculate_interval(clock, NTP_INTERVAL_LENGTH);
-       clock->cycle_last = clocksource_read(clock);
-
-       xtime.tv_sec = sec;
-       xtime.tv_nsec = 0;
+       clock = clocksource_default_clock();
+       if (clock->enable)
+               clock->enable(clock);
+       timekeeper_setup_internals(clock);
+
+       xtime.tv_sec = now.tv_sec;
+       xtime.tv_nsec = now.tv_nsec;
+       raw_time.tv_sec = 0;
+       raw_time.tv_nsec = 0;
+       if (boot.tv_sec == 0 && boot.tv_nsec == 0) {
+               boot.tv_sec = xtime.tv_sec;
+               boot.tv_nsec = xtime.tv_nsec;
+       }
        set_normalized_timespec(&wall_to_monotonic,
-               -xtime.tv_sec, -xtime.tv_nsec);
-       update_xtime_cache(0);
-       total_sleep_time = 0;
+                               -boot.tv_sec, -boot.tv_nsec);
+       total_sleep_time.tv_sec = 0;
+       total_sleep_time.tv_nsec = 0;
        write_sequnlock_irqrestore(&xtime_lock, flags);
 }
 
 /* time in seconds when suspend began */
-static unsigned long timekeeping_suspend_time;
+static struct timespec timekeeping_suspend_time;
 
 /**
  * timekeeping_resume - Resumes the generic timekeeping subsystem.
@@ -331,24 +569,23 @@ static unsigned long timekeeping_suspend_time;
 static int timekeeping_resume(struct sys_device *dev)
 {
        unsigned long flags;
-       unsigned long now = read_persistent_clock();
+       struct timespec ts;
+
+       read_persistent_clock(&ts);
 
        clocksource_resume();
 
        write_seqlock_irqsave(&xtime_lock, flags);
 
-       if (now && (now > timekeeping_suspend_time)) {
-               unsigned long sleep_length = now - timekeeping_suspend_time;
-
-               xtime.tv_sec += sleep_length;
-               wall_to_monotonic.tv_sec -= sleep_length;
-               total_sleep_time += sleep_length;
+       if (timespec_compare(&ts, &timekeeping_suspend_time) > 0) {
+               ts = timespec_sub(ts, timekeeping_suspend_time);
+               xtime = timespec_add_safe(xtime, ts);
+               wall_to_monotonic = timespec_sub(wall_to_monotonic, ts);
+               total_sleep_time = timespec_add_safe(total_sleep_time, ts);
        }
-       update_xtime_cache(0);
        /* re-base the last cycle value */
-       clock->cycle_last = 0;
-       clock->cycle_last = clocksource_read(clock);
-       clock->error = 0;
+       timekeeper.clock->cycle_last = timekeeper.clock->read(timekeeper.clock);
+       timekeeper.ntp_error = 0;
        timekeeping_suspended = 0;
        write_sequnlock_irqrestore(&xtime_lock, flags);
 
@@ -366,14 +603,15 @@ static int timekeeping_suspend(struct sys_device *dev, pm_message_t state)
 {
        unsigned long flags;
 
-       timekeeping_suspend_time = read_persistent_clock();
+       read_persistent_clock(&timekeeping_suspend_time);
 
        write_seqlock_irqsave(&xtime_lock, flags);
-       clocksource_forward_now();
+       timekeeping_forward_now();
        timekeeping_suspended = 1;
        write_sequnlock_irqrestore(&xtime_lock, flags);
 
        clockevents_notify(CLOCK_EVT_NOTIFY_SUSPEND, NULL);
+       clocksource_suspend();
 
        return 0;
 }
@@ -404,7 +642,7 @@ device_initcall(timekeeping_init_device);
  * If the error is already larger, we look ahead even further
  * to compensate for late or lost adjustments.
  */
-static __always_inline int clocksource_bigadjust(s64 error, s64 *interval,
+static __always_inline int timekeeping_bigadjust(s64 error, s64 *interval,
                                                 s64 *offset)
 {
        s64 tick_error, i;
@@ -420,7 +658,7 @@ static __always_inline int clocksource_bigadjust(s64 error, s64 *interval,
         * here.  This is tuned so that an error of about 1 msec is adjusted
         * within about 1 sec (or 2^20 nsec in 2^SHIFT_HZ ticks).
         */
-       error2 = clock->error >> (NTP_SCALE_SHIFT + 22 - 2 * SHIFT_HZ);
+       error2 = timekeeper.ntp_error >> (NTP_SCALE_SHIFT + 22 - 2 * SHIFT_HZ);
        error2 = abs(error2);
        for (look_ahead = 0; error2 > 0; look_ahead++)
                error2 >>= 2;
@@ -429,8 +667,8 @@ static __always_inline int clocksource_bigadjust(s64 error, s64 *interval,
         * Now calculate the error in (1 << look_ahead) ticks, but first
         * remove the single look ahead already included in the error.
         */
-       tick_error = tick_length >> (NTP_SCALE_SHIFT - clock->shift + 1);
-       tick_error -= clock->xtime_interval >> 1;
+       tick_error = tick_length >> (timekeeper.ntp_error_shift + 1);
+       tick_error -= timekeeper.xtime_interval >> 1;
        error = ((error - tick_error) >> look_ahead) + tick_error;
 
        /* Finally calculate the adjustment shift value.  */
@@ -455,18 +693,18 @@ static __always_inline int clocksource_bigadjust(s64 error, s64 *interval,
  * this is optimized for the most common adjustments of -1,0,1,
  * for other values we can do a bit more work.
  */
-static void clocksource_adjust(s64 offset)
+static void timekeeping_adjust(s64 offset)
 {
-       s64 error, interval = clock->cycle_interval;
+       s64 error, interval = timekeeper.cycle_interval;
        int adj;
 
-       error = clock->error >> (NTP_SCALE_SHIFT - clock->shift - 1);
+       error = timekeeper.ntp_error >> (timekeeper.ntp_error_shift - 1);
        if (error > interval) {
                error >>= 2;
                if (likely(error <= interval))
                        adj = 1;
                else
-                       adj = clocksource_bigadjust(error, &interval, &offset);
+                       adj = timekeeping_bigadjust(error, &interval, &offset);
        } else if (error < -interval) {
                error >>= 2;
                if (likely(error >= -interval)) {
@@ -474,17 +712,62 @@ static void clocksource_adjust(s64 offset)
                        interval = -interval;
                        offset = -offset;
                } else
-                       adj = clocksource_bigadjust(error, &interval, &offset);
+                       adj = timekeeping_bigadjust(error, &interval, &offset);
        } else
                return;
 
-       clock->mult += adj;
-       clock->xtime_interval += interval;
-       clock->xtime_nsec -= offset;
-       clock->error -= (interval - offset) <<
-                       (NTP_SCALE_SHIFT - clock->shift);
+       timekeeper.mult += adj;
+       timekeeper.xtime_interval += interval;
+       timekeeper.xtime_nsec -= offset;
+       timekeeper.ntp_error -= (interval - offset) <<
+                               timekeeper.ntp_error_shift;
 }
 
+
+/**
+ * logarithmic_accumulation - shifted accumulation of cycles
+ *
+ * This functions accumulates a shifted interval of cycles into
+ * into a shifted interval nanoseconds. Allows for O(log) accumulation
+ * loop.
+ *
+ * Returns the unconsumed cycles.
+ */
+static cycle_t logarithmic_accumulation(cycle_t offset, int shift)
+{
+       u64 nsecps = (u64)NSEC_PER_SEC << timekeeper.shift;
+
+       /* If the offset is smaller then a shifted interval, do nothing */
+       if (offset < timekeeper.cycle_interval<<shift)
+               return offset;
+
+       /* Accumulate one shifted interval */
+       offset -= timekeeper.cycle_interval << shift;
+       timekeeper.clock->cycle_last += timekeeper.cycle_interval << shift;
+
+       timekeeper.xtime_nsec += timekeeper.xtime_interval << shift;
+       while (timekeeper.xtime_nsec >= nsecps) {
+               timekeeper.xtime_nsec -= nsecps;
+               xtime.tv_sec++;
+               second_overflow();
+       }
+
+       /* Accumulate into raw time */
+       raw_time.tv_nsec += timekeeper.raw_interval << shift;;
+       while (raw_time.tv_nsec >= NSEC_PER_SEC) {
+               raw_time.tv_nsec -= NSEC_PER_SEC;
+               raw_time.tv_sec++;
+       }
+
+       /* Accumulate error between NTP and clock interval */
+       timekeeper.ntp_error += tick_length << shift;
+       timekeeper.ntp_error -= timekeeper.xtime_interval <<
+                               (timekeeper.ntp_error_shift + shift);
+
+       return offset;
+}
+
+
 /**
  * update_wall_time - Uses the current clocksource to increment the wall time
  *
@@ -492,53 +775,49 @@ static void clocksource_adjust(s64 offset)
  */
 void update_wall_time(void)
 {
+       struct clocksource *clock;
        cycle_t offset;
+       int shift = 0, maxshift;
 
        /* Make sure we're fully resumed: */
        if (unlikely(timekeeping_suspended))
                return;
 
+       clock = timekeeper.clock;
 #ifdef CONFIG_GENERIC_TIME
-       offset = (clocksource_read(clock) - clock->cycle_last) & clock->mask;
+       offset = (clock->read(clock) - clock->cycle_last) & clock->mask;
 #else
-       offset = clock->cycle_interval;
+       offset = timekeeper.cycle_interval;
 #endif
-       clock->xtime_nsec = (s64)xtime.tv_nsec << clock->shift;
+       timekeeper.xtime_nsec = (s64)xtime.tv_nsec << timekeeper.shift;
 
-       /* normally this loop will run just once, however in the
-        * case of lost or late ticks, it will accumulate correctly.
+       /*
+        * With NO_HZ we may have to accumulate many cycle_intervals
+        * (think "ticks") worth of time at once. To do this efficiently,
+        * we calculate the largest doubling multiple of cycle_intervals
+        * that is smaller then the offset. We then accumulate that
+        * chunk in one go, and then try to consume the next smaller
+        * doubled multiple.
         */
-       while (offset >= clock->cycle_interval) {
-               /* accumulate one interval */
-               offset -= clock->cycle_interval;
-               clock->cycle_last += clock->cycle_interval;
-
-               clock->xtime_nsec += clock->xtime_interval;
-               if (clock->xtime_nsec >= (u64)NSEC_PER_SEC << clock->shift) {
-                       clock->xtime_nsec -= (u64)NSEC_PER_SEC << clock->shift;
-                       xtime.tv_sec++;
-                       second_overflow();
-               }
-
-               clock->raw_time.tv_nsec += clock->raw_interval;
-               if (clock->raw_time.tv_nsec >= NSEC_PER_SEC) {
-                       clock->raw_time.tv_nsec -= NSEC_PER_SEC;
-                       clock->raw_time.tv_sec++;
-               }
-
-               /* accumulate error between NTP and clock interval */
-               clock->error += tick_length;
-               clock->error -= clock->xtime_interval << (NTP_SCALE_SHIFT - clock->shift);
+       shift = ilog2(offset) - ilog2(timekeeper.cycle_interval);
+       shift = max(0, shift);
+       /* Bound shift to one less then what overflows tick_length */
+       maxshift = (8*sizeof(tick_length) - (ilog2(tick_length)+1)) - 1;
+       shift = min(shift, maxshift);
+       while (offset >= timekeeper.cycle_interval) {
+               offset = logarithmic_accumulation(offset, shift);
+               if(offset < timekeeper.cycle_interval<<shift)
+                       shift--;
        }
 
        /* correct the clock when NTP error is too big */
-       clocksource_adjust(offset);
+       timekeeping_adjust(offset);
 
        /*
         * Since in the loop above, we accumulate any amount of time
         * in xtime_nsec over a second into xtime.tv_sec, its possible for
         * xtime_nsec to be fairly small after the loop. Further, if we're
-        * slightly speeding the clocksource up in clocksource_adjust(),
+        * slightly speeding the clocksource up in timekeeping_adjust(),
         * its possible the required corrective factor to xtime_nsec could
         * cause it to underflow.
         *
@@ -550,24 +829,34 @@ void update_wall_time(void)
         * We'll correct this error next time through this function, when
         * xtime_nsec is not as small.
         */
-       if (unlikely((s64)clock->xtime_nsec < 0)) {
-               s64 neg = -(s64)clock->xtime_nsec;
-               clock->xtime_nsec = 0;
-               clock->error += neg << (NTP_SCALE_SHIFT - clock->shift);
+       if (unlikely((s64)timekeeper.xtime_nsec < 0)) {
+               s64 neg = -(s64)timekeeper.xtime_nsec;
+               timekeeper.xtime_nsec = 0;
+               timekeeper.ntp_error += neg << timekeeper.ntp_error_shift;
        }
 
-       /* store full nanoseconds into xtime after rounding it up and
+
+       /*
+        * Store full nanoseconds into xtime after rounding it up and
         * add the remainder to the error difference.
         */
-       xtime.tv_nsec = ((s64)clock->xtime_nsec >> clock->shift) + 1;
-       clock->xtime_nsec -= (s64)xtime.tv_nsec << clock->shift;
-       clock->error += clock->xtime_nsec << (NTP_SCALE_SHIFT - clock->shift);
+       xtime.tv_nsec = ((s64) timekeeper.xtime_nsec >> timekeeper.shift) + 1;
+       timekeeper.xtime_nsec -= (s64) xtime.tv_nsec << timekeeper.shift;
+       timekeeper.ntp_error += timekeeper.xtime_nsec <<
+                               timekeeper.ntp_error_shift;
 
-       update_xtime_cache(cyc2ns(clock, offset));
+       /*
+        * Finally, make sure that after the rounding
+        * xtime.tv_nsec isn't larger then NSEC_PER_SEC
+        */
+       if (unlikely(xtime.tv_nsec >= NSEC_PER_SEC)) {
+               xtime.tv_nsec -= NSEC_PER_SEC;
+               xtime.tv_sec++;
+               second_overflow();
+       }
 
        /* check to see if there is a new clocksource to use */
-       change_clocksource();
-       update_vsyscall(&xtime, clock);
+       update_vsyscall(&xtime, timekeeper.clock, timekeeper.mult);
 }
 
 /**
@@ -583,10 +872,14 @@ void update_wall_time(void)
  */
 void getboottime(struct timespec *ts)
 {
-       set_normalized_timespec(ts,
-               - (wall_to_monotonic.tv_sec + total_sleep_time),
-               - wall_to_monotonic.tv_nsec);
+       struct timespec boottime = {
+               .tv_sec = wall_to_monotonic.tv_sec + total_sleep_time.tv_sec,
+               .tv_nsec = wall_to_monotonic.tv_nsec + total_sleep_time.tv_nsec
+       };
+
+       set_normalized_timespec(ts, -boottime.tv_sec, -boottime.tv_nsec);
 }
+EXPORT_SYMBOL_GPL(getboottime);
 
 /**
  * monotonic_to_bootbased - Convert the monotonic time to boot based.
@@ -594,15 +887,20 @@ void getboottime(struct timespec *ts)
  */
 void monotonic_to_bootbased(struct timespec *ts)
 {
-       ts->tv_sec += total_sleep_time;
+       *ts = timespec_add_safe(*ts, total_sleep_time);
 }
+EXPORT_SYMBOL_GPL(monotonic_to_bootbased);
 
 unsigned long get_seconds(void)
 {
-       return xtime_cache.tv_sec;
+       return xtime.tv_sec;
 }
 EXPORT_SYMBOL(get_seconds);
 
+struct timespec __current_kernel_time(void)
+{
+       return xtime;
+}
 
 struct timespec current_kernel_time(void)
 {
@@ -612,9 +910,26 @@ struct timespec current_kernel_time(void)
        do {
                seq = read_seqbegin(&xtime_lock);
 
-               now = xtime_cache;
+               now = xtime;
        } while (read_seqretry(&xtime_lock, seq));
 
        return now;
 }
 EXPORT_SYMBOL(current_kernel_time);
+
+struct timespec get_monotonic_coarse(void)
+{
+       struct timespec now, mono;
+       unsigned long seq;
+
+       do {
+               seq = read_seqbegin(&xtime_lock);
+
+               now = xtime;
+               mono = wall_to_monotonic;
+       } while (read_seqretry(&xtime_lock, seq));
+
+       set_normalized_timespec(&now, now.tv_sec + mono.tv_sec,
+                               now.tv_nsec + mono.tv_nsec);
+       return now;
+}